Does Safegraph correlate with COVID19 cases?
Ohio – confirmed – FPCA
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.2481056 1.5692590 -0.158 0.87438
## cbgam2v1.l1 -0.1655405 0.0238200 -6.950 3.80e-12 ***
## cbgam2v1.l2 -1.2390660 0.0998974 -12.403 < 2e-16 ***
## cbgam2v1.l3 1.0228713 0.2730512 3.746 0.00018 ***
## cbgam2v2.l1 -0.2167554 0.0265919 -8.151 3.87e-16 ***
## cbgam2v2.l2 -1.4623201 0.1098598 -13.311 < 2e-16 ***
## cbgam2v2.l3 1.2753285 0.3146705 4.053 5.08e-05 ***
## cbgam2v3.l1 -0.1976064 0.0274890 -7.189 6.83e-13 ***
## cbgam2v3.l2 -1.4713590 0.1100680 -13.368 < 2e-16 ***
## cbgam2v3.l3 1.2589822 0.3141729 4.007 6.17e-05 ***
## cbgam2v4.l1 -0.3007896 0.0281447 -10.687 < 2e-16 ***
## cbgam2v4.l2 -1.4281939 0.1107210 -12.899 < 2e-16 ***
## cbgam2v4.l3 1.4020933 0.3135844 4.471 7.83e-06 ***
## cbgam2v5.l1 -0.3202169 0.0297952 -10.747 < 2e-16 ***
## cbgam2v5.l2 -1.4732199 0.1102811 -13.359 < 2e-16 ***
## cbgam2v5.l3 1.4714920 0.3140096 4.686 2.81e-06 ***
## cbgam2v6.l1 -0.4566587 0.0591895 -7.715 1.28e-14 ***
## cbgam2v6.l2 -1.9927339 0.1295657 -15.380 < 2e-16 ***
## cbgam2v6.l3 0.9684784 0.3218737 3.009 0.00263 **
## density -0.0007044 0.0004548 -1.549 0.12146
## ccvi_quintile20-40 -0.0025860 0.2708127 -0.010 0.99238
## ccvi_quintile40-60 -0.4260916 0.2757714 -1.545 0.12234
## ccvi_quintile60-80 -0.1037217 0.2749235 -0.377 0.70597
## ccvi_quintile80-100 -0.6191789 0.2901762 -2.134 0.03287 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.845 8.985 174.28 <2e-16 ***
## s(time,county) 317.116 428.000 15.52 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.859 Deviance explained = 83.6%
## -REML = 24068 Scale est. = 6.1145 n = 16530Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 18568
## groups: 88
## range: -4.949771 to 5.244575
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| -0.05 | [-0.23, 0.10] |
Ohio – confirmed – not_at_home_device_count_change
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -9.4543073 0.2991433 -31.605 < 2e-16 ***
## cbgam2v1.l1 0.0449994 0.0123990 3.629 0.000285 ***
## cbgam2v1.l2 -0.0053036 0.0129298 -0.410 0.681676
## cbgam2v1.l3 -0.0077687 0.0208452 -0.373 0.709387
## cbgam2v2.l1 0.0650969 0.0146151 4.454 8.48e-06 ***
## cbgam2v2.l2 0.0198566 0.0150739 1.317 0.187763
## cbgam2v2.l3 -0.0311140 0.0232453 -1.339 0.180750
## cbgam2v3.l1 0.0799138 0.0140899 5.672 1.44e-08 ***
## cbgam2v3.l2 0.0167833 0.0150622 1.114 0.265181
## cbgam2v3.l3 -0.0454964 0.0225752 -2.015 0.043887 *
## cbgam2v4.l1 0.0572146 0.0149341 3.831 0.000128 ***
## cbgam2v4.l2 0.0258286 0.0159658 1.618 0.105737
## cbgam2v4.l3 -0.0247039 0.0229523 -1.076 0.281802
## cbgam2v5.l1 0.1063144 0.0165658 6.418 1.42e-10 ***
## cbgam2v5.l2 0.0525868 0.0175319 2.999 0.002708 **
## cbgam2v5.l3 -0.0344500 0.0245444 -1.404 0.160463
## cbgam2v6.l1 0.0917621 0.0228128 4.022 5.79e-05 ***
## cbgam2v6.l2 0.0328208 0.0235597 1.393 0.163612
## cbgam2v6.l3 -0.0337893 0.0361665 -0.934 0.350178
## density -0.0005019 0.0004403 -1.140 0.254387
## ccvi_quintile20-40 -0.0091679 0.2629404 -0.035 0.972186
## ccvi_quintile40-60 -0.3970620 0.2677627 -1.483 0.138124
## ccvi_quintile60-80 -0.1578364 0.2668291 -0.592 0.554176
## ccvi_quintile80-100 -0.5722001 0.2810734 -2.036 0.041790 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.859 8.987 204.9 <2e-16 ***
## s(time,county) 304.254 428.000 3377.7 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.812 Deviance explained = 81.7%
## -REML = 24925 Scale est. = 7.2321 n = 16530Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 18568
## groups: 88
## range: -7.013312 to 8.403648
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| -0.02 | [-0.10, 0.04] |
Missouri – confirmed – FPCA
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -7.9929480 1.0618633 -7.527 5.38e-14 ***
## cbgam2v1.l1 0.0497405 0.0356118 1.397 0.1625
## cbgam2v1.l2 -0.0286861 0.0598674 -0.479 0.6318
## cbgam2v1.l3 -0.1182001 0.0913424 -1.294 0.1957
## cbgam2v2.l1 -0.0247296 0.0366514 -0.675 0.4999
## cbgam2v2.l2 -0.0633285 0.0650804 -0.973 0.3305
## cbgam2v2.l3 -0.1250820 0.1007224 -1.242 0.2143
## cbgam2v3.l1 -0.0347312 0.0370444 -0.938 0.3485
## cbgam2v3.l2 -0.0697899 0.0654490 -1.066 0.2863
## cbgam2v3.l3 -0.1624370 0.1004607 -1.617 0.1059
## cbgam2v4.l1 -0.0036520 0.0374538 -0.098 0.9223
## cbgam2v4.l2 0.0020219 0.0662365 0.031 0.9756
## cbgam2v4.l3 -0.0038659 0.1008362 -0.038 0.9694
## cbgam2v5.l1 -0.0138600 0.0390379 -0.355 0.7226
## cbgam2v5.l2 -0.0030113 0.0661258 -0.046 0.9637
## cbgam2v5.l3 0.0565347 0.1007875 0.561 0.5749
## cbgam2v6.l1 0.0586247 0.0551625 1.063 0.2879
## cbgam2v6.l2 0.1488674 0.0843854 1.764 0.0777 .
## cbgam2v6.l3 0.2493528 0.1256891 1.984 0.0473 *
## density -0.0002902 0.0002930 -0.990 0.3221
## ccvi_quintile20-40 0.0335494 0.2073085 0.162 0.8714
## ccvi_quintile40-60 0.2513591 0.2127677 1.181 0.2375
## ccvi_quintile60-80 0.2097265 0.2159804 0.971 0.3315
## ccvi_quintile80-100 0.1089889 0.2184387 0.499 0.6178
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.83 8.985 130.6 <2e-16 ***
## s(time,county) 319.88 563.000 10452.5 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.763 Deviance explained = 55.8%
## -REML = 35194 Scale est. = 10.3 n = 22074Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 24687
## groups: 115
## range: -3.915483 to 6.277683
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| 0.36 | [0.23, 0.47] |
Missouri – confirmed – not_at_home_device_count_change
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -8.6785296 0.2714111 -31.976 <2e-16 ***
## cbgam2v1.l1 0.0118280 0.0097328 1.215 0.224
## cbgam2v1.l2 0.0001098 0.0114555 0.010 0.992
## cbgam2v1.l3 0.0002999 0.0170405 0.018 0.986
## cbgam2v2.l1 0.0056960 0.0111685 0.510 0.610
## cbgam2v2.l2 -0.0086949 0.0141383 -0.615 0.539
## cbgam2v2.l3 -0.0111134 0.0226416 -0.491 0.624
## cbgam2v3.l1 -0.0077217 0.0110971 -0.696 0.487
## cbgam2v3.l2 -0.0180543 0.0144857 -1.246 0.213
## cbgam2v3.l3 -0.0295217 0.0232215 -1.271 0.204
## cbgam2v4.l1 -0.0081136 0.0119748 -0.678 0.498
## cbgam2v4.l2 -0.0096718 0.0152620 -0.634 0.526
## cbgam2v4.l3 -0.0382553 0.0232982 -1.642 0.101
## cbgam2v5.l1 -0.0074997 0.0142062 -0.528 0.598
## cbgam2v5.l2 -0.0190017 0.0180209 -1.054 0.292
## cbgam2v5.l3 -0.0392191 0.0255709 -1.534 0.125
## cbgam2v6.l1 0.0012517 0.0216489 0.058 0.954
## cbgam2v6.l2 -0.0233125 0.0266764 -0.874 0.382
## cbgam2v6.l3 -0.0463473 0.0373738 -1.240 0.215
## density -0.0002025 0.0002758 -0.734 0.463
## ccvi_quintile20-40 0.0573403 0.1971000 0.291 0.771
## ccvi_quintile40-60 0.2568785 0.2023151 1.270 0.204
## ccvi_quintile60-80 0.2355733 0.2050587 1.149 0.251
## ccvi_quintile80-100 0.1609694 0.2075387 0.776 0.438
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.827 8.985 147.6 <2e-16 ***
## s(time,county) 313.700 563.000 30500.1 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.757 Deviance explained = 55%
## -REML = 35377 Scale est. = 10.457 n = 22074Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 24687
## groups: 115
## range: -5.26114 to 4.783962
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| 0.08 | [0.02, 0.14] |
South Carolina – confirmed – FPCA
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -4.9710735 3.1244804 -1.591 0.111642
## cbgam2v1.l1 -0.3651368 0.1943512 -1.879 0.060311 .
## cbgam2v1.l2 -0.1442688 0.1977649 -0.729 0.465717
## cbgam2v1.l3 0.3423535 0.2575586 1.329 0.183807
## cbgam2v2.l1 -0.4369763 0.2047065 -2.135 0.032816 *
## cbgam2v2.l2 -0.1245427 0.2098121 -0.594 0.552800
## cbgam2v2.l3 0.1806986 0.2722340 0.664 0.506859
## cbgam2v3.l1 -0.3563536 0.2090563 -1.705 0.088306 .
## cbgam2v3.l2 -0.0833101 0.2151986 -0.387 0.698668
## cbgam2v3.l3 0.0834321 0.2761956 0.302 0.762601
## cbgam2v4.l1 -0.3699696 0.2080692 -1.778 0.075419 .
## cbgam2v4.l2 -0.1014235 0.2135369 -0.475 0.634820
## cbgam2v4.l3 0.0855269 0.2741439 0.312 0.755064
## cbgam2v5.l1 -0.1885545 0.2101281 -0.897 0.369566
## cbgam2v5.l2 -0.1744950 0.2233516 -0.781 0.434671
## cbgam2v5.l3 0.0034195 0.2953735 0.012 0.990763
## cbgam2v6.l1 -0.3223380 0.2266019 -1.422 0.154919
## cbgam2v6.l2 -1.1599821 0.3370980 -3.441 0.000582 ***
## cbgam2v6.l3 -0.1002039 0.5366240 -0.187 0.851876
## density -0.0007326 0.0039610 -0.185 0.853277
## ccvi_quintile20-40 -0.1377679 0.5404441 -0.255 0.798794
## ccvi_quintile40-60 0.0058617 0.5900717 0.010 0.992074
## ccvi_quintile60-80 -0.4319927 0.6516090 -0.663 0.507371
## ccvi_quintile80-100 -0.7240565 0.6607208 -1.096 0.273170
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.74 8.96 67.00 <2e-16 ***
## s(time,county) 166.76 218.00 57.98 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.69 Deviance explained = 51.5%
## -REML = 14248 Scale est. = 11.412 n = 9360Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 10534
## groups: 46
## range: -2.666375 to 3.187262
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| -0.28 | [-0.80, 0.09] |
South Carolina – confirmed – not_at_home_device_count_change
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -7.3585636 0.7044556 -10.446 < 2e-16 ***
## cbgam2v1.l1 -0.0007781 0.0156873 -0.050 0.960441
## cbgam2v1.l2 -0.0221420 0.0199778 -1.108 0.267748
## cbgam2v1.l3 -0.0769038 0.0297273 -2.587 0.009698 **
## cbgam2v2.l1 -0.0170498 0.0146547 -1.163 0.244682
## cbgam2v2.l2 -0.0358032 0.0203529 -1.759 0.078590 .
## cbgam2v2.l3 -0.0237120 0.0317925 -0.746 0.455787
## cbgam2v3.l1 -0.0245017 0.0141348 -1.733 0.083054 .
## cbgam2v3.l2 -0.0453016 0.0205697 -2.202 0.027666 *
## cbgam2v3.l3 -0.0875826 0.0320879 -2.729 0.006356 **
## cbgam2v4.l1 -0.0470156 0.0161181 -2.917 0.003543 **
## cbgam2v4.l2 -0.0759029 0.0225900 -3.360 0.000783 ***
## cbgam2v4.l3 -0.0940610 0.0328642 -2.862 0.004218 **
## cbgam2v5.l1 -0.0536121 0.0187174 -2.864 0.004189 **
## cbgam2v5.l2 -0.0819148 0.0254076 -3.224 0.001268 **
## cbgam2v5.l3 -0.1169673 0.0361538 -3.235 0.001220 **
## cbgam2v6.l1 -0.0201357 0.0258199 -0.780 0.435498
## cbgam2v6.l2 -0.0635720 0.0348997 -1.822 0.068554 .
## cbgam2v6.l3 -0.1115645 0.0507456 -2.199 0.027938 *
## density -0.0020274 0.0038325 -0.529 0.596821
## ccvi_quintile20-40 -0.1721584 0.5236662 -0.329 0.742348
## ccvi_quintile40-60 -0.0310760 0.5724979 -0.054 0.956712
## ccvi_quintile60-80 -0.4288778 0.6317362 -0.679 0.497226
## ccvi_quintile80-100 -0.7733048 0.6405953 -1.207 0.227399
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.844 8.981 84.68 <2e-16 ***
## s(time,county) 166.336 218.000 107719.99 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.673 Deviance explained = 49.2%
## -REML = 14441 Scale est. = 11.263 n = 9360Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 10534
## groups: 46
## range: -5.293829 to 4.110733
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| -0.02 | [-0.14, 0.09] |
Indiana – confirmed – FPCA
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -7.3008451 0.7595928 -9.612 < 2e-16 ***
## cbgam2v1.l1 -0.0067489 0.0328058 -0.206 0.837010
## cbgam2v1.l2 -0.0387647 0.0624688 -0.621 0.534907
## cbgam2v1.l3 -0.2073509 0.0885050 -2.343 0.019150 *
## cbgam2v2.l1 -0.1298959 0.0318574 -4.077 4.57e-05 ***
## cbgam2v2.l2 0.0671417 0.0656669 1.022 0.306578
## cbgam2v2.l3 -0.2694037 0.0952076 -2.830 0.004665 **
## cbgam2v3.l1 -0.0975700 0.0322650 -3.024 0.002498 **
## cbgam2v3.l2 0.0706172 0.0661173 1.068 0.285508
## cbgam2v3.l3 -0.2454733 0.0952109 -2.578 0.009940 **
## cbgam2v4.l1 -0.1531421 0.0326987 -4.683 2.84e-06 ***
## cbgam2v4.l2 0.0018073 0.0667207 0.027 0.978390
## cbgam2v4.l3 -0.1665106 0.0947528 -1.757 0.078881 .
## cbgam2v5.l1 -0.1655653 0.0341055 -4.855 1.22e-06 ***
## cbgam2v5.l2 -0.1174962 0.0660370 -1.779 0.075216 .
## cbgam2v5.l3 -0.0645839 0.0949800 -0.680 0.496530
## cbgam2v6.l1 -0.2264953 0.0676473 -3.348 0.000815 ***
## cbgam2v6.l2 -0.0684587 0.0941513 -0.727 0.467166
## cbgam2v6.l3 0.2005718 0.1212982 1.654 0.098238 .
## density -0.0004831 0.0007048 -0.686 0.493029
## ccvi_quintile20-40 -0.1092601 0.2661421 -0.411 0.681420
## ccvi_quintile40-60 0.1219913 0.2636666 0.463 0.643605
## ccvi_quintile60-80 -0.1207748 0.2731773 -0.442 0.658414
## ccvi_quintile80-100 -0.3734846 0.2804778 -1.332 0.183008
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.911 8.992 194.3 <2e-16 ***
## s(time,county) 339.389 448.000 281.9 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.926 Deviance explained = 83.9%
## -REML = 22259 Scale est. = 4.4652 n = 18018Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 20148
## groups: 92
## range: -4.077463 to 3.234633
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| -0.53 | [-0.74, -0.34] |
Indiana – confirmed – not_at_home_device_count_change
##
## Family: quasipoisson
## Link function: log
##
## Formula:
## daily_count ~ cbgam2 + density + ccvi_quintile + s(time, bs = "tp") +
## s(time, county, bs = c("fs"), k = 5, m = 2) + offset(log(population))
##
## Parametric coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -8.8068680 0.3076318 -28.628 < 2e-16 ***
## cbgam2v1.l1 0.0409080 0.0126175 3.242 0.001188 **
## cbgam2v1.l2 0.0143243 0.0134283 1.067 0.286112
## cbgam2v1.l3 -0.0159244 0.0197479 -0.806 0.420033
## cbgam2v2.l1 0.0436933 0.0129840 3.365 0.000767 ***
## cbgam2v2.l2 0.0154499 0.0141353 1.093 0.274408
## cbgam2v2.l3 -0.0893255 0.0208115 -4.292 1.78e-05 ***
## cbgam2v3.l1 0.0371242 0.0129420 2.869 0.004129 **
## cbgam2v3.l2 0.0045478 0.0142401 0.319 0.749454
## cbgam2v3.l3 -0.1101936 0.0205670 -5.358 8.53e-08 ***
## cbgam2v4.l1 0.0218108 0.0138549 1.574 0.115453
## cbgam2v4.l2 0.0022238 0.0150243 0.148 0.882336
## cbgam2v4.l3 -0.1137870 0.0207644 -5.480 4.31e-08 ***
## cbgam2v5.l1 0.0459453 0.0154664 2.971 0.002976 **
## cbgam2v5.l2 0.0228195 0.0168493 1.354 0.175648
## cbgam2v5.l3 -0.0827534 0.0228805 -3.617 0.000299 ***
## cbgam2v6.l1 -0.0011756 0.0218131 -0.054 0.957021
## cbgam2v6.l2 -0.0100365 0.0225304 -0.445 0.655990
## cbgam2v6.l3 -0.1235514 0.0329734 -3.747 0.000180 ***
## density -0.0002954 0.0007174 -0.412 0.680503
## ccvi_quintile20-40 -0.0504230 0.2713492 -0.186 0.852586
## ccvi_quintile40-60 0.1588427 0.2692094 0.590 0.555175
## ccvi_quintile60-80 0.0688566 0.2779766 0.248 0.804364
## ccvi_quintile80-100 -0.2695863 0.2857737 -0.943 0.345512
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Approximate significance of smooth terms:
## edf Ref.df F p-value
## s(time) 8.903 8.991 332.9 <2e-16 ***
## s(time,county) 338.173 448.000 9820.9 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## R-sq.(adj) = 0.914 Deviance explained = 83%
## -REML = 22743 Scale est. = 4.708 n = 18018Time effect
This show the effect of calendar time on incidence of cases/deaths, global and interaction with time:
Time by county interaction
PCA by CCVI quintile interaction
3d plot
Contour plot
By Z-score
By Lag
Overall
crossbasis summary
## CROSSBASIS FUNCTIONS
## observations: 20148
## groups: 92
## range: -5.348223 to 4.980483
## lag period: 0 21
## total df: 18
##
## BASIS FOR VAR:
## fun: cr
## df: 6
## knots: -3 -2 -1 0 1 2 3 ...
## intercept: FALSE
## fx: FALSE
## S: 9.876923 -9.507692 4.153846 -1.107692 0.2769231 -0.04615385 -9.507692 ...
##
## BASIS FOR LAG:
## fun: cr
## df: 3
## knots: 0 7 14 21
## intercept: FALSE
## fx: FALSE
## S: 0.02798834 -0.0244898 0.006997085 -0.0244898 0.02798834 -0.01049563 0.006997085 ...Attributable Fractions (AF)
| AF | CI |
|---|---|
| 0.04 | [-0.03, 0.10] |
Likelihood Ratio Tests
A lower Resid. Dev indicates a better fit. The first row of each table is for single Metric, the second row is for PCA.
[[1]]
| Resid. Df | Resid. Dev | Df | Deviance | Pr(>Chi) |
|---|---|---|---|---|
| 16150.05 | 112024.36 | NA | NA | NA |
| 16138.62 | 99950.19 | 11.42611 | 12074.17 | 0 |
[[2]]
| Resid. Df | Resid. Dev | Df | Deviance | Pr(>Chi) |
|---|---|---|---|---|
| 21673.97 | 187439.6 | NA | NA | NA |
| 21666.96 | 183785.0 | 7.014844 | 3654.553 | 0 |
[[3]]
| Resid. Df | Resid. Dev | Df | Deviance | Pr(>Chi) |
|---|---|---|---|---|
| 9142.229 | 67757.48 | NA | NA | NA |
| 9139.455 | 64724.88 | 2.773504 | 3032.599 | 0 |
[[4]]
| Resid. Df | Resid. Dev | Df | Deviance | Pr(>Chi) |
|---|---|---|---|---|
| 17602.82 | 74436.15 | NA | NA | NA |
| 17602.24 | 70416.28 | 0.5821207 | 4019.869 | 0 |